Cell type identification of the single cells of Drosophila optic lobe based on their specific markers
To identify the cells corresponding to distinct subtypes of glia and neurons, the expression profile of the sequenced single cells by Konstantinides et al.  was analyzed based on their previously-reported marker genes (Fig. 1A). According to the expression profile of 55000 sequenced single cells, the cells with the highest expression of each cell type-specific marker were selected and classified. A total of forty-two cells were identified and classified into seven cell types. The heatmap of gene expression for these selected cells confirmed the unique expression pattern of each marker in the corresponding cell types (Fig. 1B). CG4797 and gem are differentially expressed in PNG, and SPG were distinguished by the three markers Mdr65, Moody and Gli.
Other glial cells were successfully distinguished by single marker; Cyp4g15, CG9657, CG34340 and Eaat1. Moreover, these genes could individually separate CG, EGN, EGT and ALG, respectively. Neuronal cells were determined based on the highest expression of nSyb and elav genes (Fig. 1B).
Hierarchical clustering of Drosophila neural cell subtypes based on their similarities in transcriptome
As shown above, different types of Drosophila neural cells could be classified based on some specific marker genes (Fig. 1). In order to know whether the classified cells in each category have the same expression profile or not, correlation analysis between the selected cells was carried out based on their transcriptome analysis. Results showed that most of the cells expressing the same genetic marker(s) tend to be clustered together, suggesting their similar transcriptome and accuracy of the identification of cell types (Fig. 2A). However, the cells with the highest expression of SPG markers and some other cells could not be classified in a single cluster suggesting their transcriptome dissimilarity or partial insufficiency of the available genetic markers for clustering (e.g. expression heterogeneity of the SPG markers: Mdr65, Moody and Gli).
With at least 3 of 6 similar cells in each category, differential expression analysis was performed in order to find the genes with specific expression in each cell type. While the identified PNG were positive for the expression of their two known genetic markers (CG4797 and gem) (Fig. 1B), there were two sub-clusters of PNG (PNG-α and PNG-β) in our transcriptome analysis which suggests the presence of a novel subtype for PNG (Fig. 2A). This analysis was also performed on another set of 42 cells of 55,000 cells and the same results were obtained (Supplemental file S1), which could support the accuracy of cell type characterization and clustering.
Also, the list of differentially expressed genes for each cell type in addition to differentially expressed genes between the two subtypes of PNG is presented in supplemental file S2.
Using the transcriptome data of the studied cell types and their clustering results, we assigned the degree of differences between them as a dendrogram (Fig. 2B). Accordingly, the Venn diagram was used to illustrate the number of unique and shared differentially-expressed gene for each cell type. The extent of unique differentially expressed genes for each cell type is as follows: 98.33% (59 of 60) for PNG, 100% (1 of 1) for SPG, 90% (36 of 40) for CG, 60% (3 of 5) for EG, 78.95 (75 of 95) for ALG and 96.39 (454 of 471) for Neurons (Fig. 2C).
Identification of deathstar as a gene with specific enrichment in astrocyte-like glia of D. melanogaster
After cell type identification, we asked whether specific genes can also characterize the identified cell types. To identify the genes capable of distinguishing different cell types, we performed differential expression analysis in order to compare the transcriptome of the cluster of interest with that of all other cell types. The differentially expressed genes (DEGs) for each cell type are presented in Supplemental file S2.
Among all the clustered cells, ALG could be sufficiently discriminated by a single uncharacterized gene named CG11000 (here assigned as deathstar). Expression level analysis of this gene using the RNA-seq profile of single cells in Drosophila optic lobe  demonstrated its significant enrichment in 12 cells of all 84 cells, which corresponds to ALG (Fig. 3A). In the same RNA-seq expression data (accession number: GSE103771), a significant positive correlation (R2=0.6637; p-value = 0.0001) was observed between the expression level of the candidate gene (deathstar) and that of Eaat1, which previously was reported as ALG-specific marker  (Fig. 3B). In addition, the expression pattern of deathstar and Eaat1 across developmental time of D. melanogaster, derived from a different RNA-seq data (accession number: GSE107049) showed a correlation between these two genes (Fig. 3C and 3D), supporting the observed expression specificity of deathstar gene in ALG of D. melanogaster.
To corroborate the assigned specificity of deathstar, we tested whether the promoter-GAL4 lines of this gene could label ALG in CNS of D. melanogaster. Consistently, data showed that deathstar (Red signals) is mostly co-labelled with the ALG marker Eaat-1 (Green signals), in specific bilateral parts of Drosophila optic lobe and VNC. Both male and female flies showed the same pattern of deathstar expression in Drosophila CNS (Fig. 4A and 4B). However, the signals of deathstar gene in two cells of Drosophila VNC were not overlapped with the expression of Eaat1 gene. This experiment was performed twice for males and females, and the same results were obtained (Supplemental file S3).
Sex-biased developmental effect of deathstar gene in D. melanogaster
We crossed GAL4 strains that are expressed in each neural cell types of D. melanogaster with the deathstar-dsRNA to knockdown deathstar protein level in each neural cell types. Our results showed that most of the males expressing deathstar-RNAi in all neural cell types are lethal during development. However, the females with the same genotype developed normally to adult. Both males and females in the control group (harbouring deathstar-RNAi without GAL4 expression) showed normal development (Fig. 5A). This result suggests that the deathstar gene may have a developmental function exclusively in male flies. However, such developmental effect was not cell type-specific. As the control of this experiment, the Drosophila CG15765 gene was examined in the same procedure, and a standard ratio of males over females (~1) was obtained for all genotypes (Fig. 5B). The exact number of the progenies of the corresponding crosses are summarized in the supplemental file S4.
Knockdown of deathstar gene affects locomotion activity of D. melanogaster
It has been previously reported that ALG cells that cover specific territories of Drosophila CNS have significant effects on locomotion behaviours . To investigate the role of deathstar gene in locomotion behaviour in D. melanogaster, we knocked down the deathstar protein level using RNAi in different cell types of CNS. Then we performed the climbing assay. Results showed that the flies that specifically express deathstar RNAi in astrocytes show significant defects in climbing ability (Fig. 6A), suggesting the crucial role of this gene in this cell type. Interestingly, we found that the knockdown of deathstar protein does not affect the climbing activity in female flies (Fig. 6B). When we knocked down the deathstar in EGN and SPG, however, female flies showed severe defects in climbing activity (Fig. 6B). These data suggest that the expression of deathstar in ALG cells is crucial only in males, not females regarding the climbing activity.
To test whether the behavioural locomotive effect of deathstar gene is altered during the ageing of D. melanogaster, the climbing assay was performed on the flies with different ages. Results showed that the older flies expressing deathstar-RNAi in all neural cell types (three- and four-week-old flies) exhibited significant climbing deficiency (Fig. 6C). However, when we tested the same genotype of flies in the one- and two-week stages, the climbing defects appeared in ALG-specific manner (Fig. 6C). The climbing defects were not observed in the flies with neuron-specific suppression of deathstar gene, suggesting the climbing defects related to deathstar is glial-specific in old flies (Fig. 6C). Unlike the male flies, both young and old females did not exhibit any meaningful pattern of climbing defects (Fig. 6D). All these data suggest that the expression of deathstar in male ALG cells is crucial for climbing ability in the early adult stage.
Astrocyte-specific suppression of deathstar gene shortens lifespan in D. melanogaster
To test the possible function of deathstar gene in D. melanogaster lifespan, we knocked down deathstar in distinct CNS cell types (CG, PNG, ALG, EGN, EGT, SPG, Glia and Neurons) and measured their life span for 40~50 days. We observed a significant reduction in lifespan for the flies that had ALG-specific expression of deathstar RNAi, supporting the specific effect of the deathstar gene in ALG on the life span of D. melanogaster (Figure 7A). However, in female flies such reducing effect of deathstar RNAi was observed in other cell types (EGN, EGT, PNG and neurons) at lower level than that of ALG. The ALG-specific effect of deathstar RNAi on lifespan was observed in both male and female flies (Fig. 7A and 7B), suggesting that the role of deathstar on the lifespan is not sexually dimorphic.